The form that works
An image describes a moment, a clip describes a change. Write what moves, how, the camera and the rhythm.Before / after
- Weak
- Improved
@Name), their portrait serves as the start image if you don’t attach one. Describe their action, never their appearance.
Stitching clips together
Ask the assistant to edit the session’s clips: “stitch the clips together”, “edit them with the music”. It chains them in the session’s order (or the one you give) with a hard cut, or a crossfade if you mention a fade or a soft transition. A music track or a voice from the session can go underneath: either alone, or mixed with the clips’ sound. This editing is free.Common mistakes
- Redescribing the image instead of the movement: say what changes.
- Too many actions in a single clip: a short clip holds one gesture or one camera move, not a whole scene.
- Forgetting the camera: without an instruction, it often stays frozen.
- Asking for a length beyond what the engine allows: the maximum length varies by engine, look at the range shown.
- An end image requested from an engine that doesn’t handle it: choose an engine that accepts it.
Settings that go with it
- Engine: maximum length, end image, native sound, number of references vary. See Choosing an engine.
- Duration, Format, Resolution: the price is counted per second, so use a short length to test.
- Sound: on or off when the engine offers it.
- On an image from the session, “Use as start image” places it in the request.
- Test short before launching long: see Spending credits wisely.